DE4329759A1 - Method for controlling a process - Google Patents

Abstract

The invention relates to a method for controlling a process which can assume different power levels, has a different dynamic behaviour at these power levels, and which forms the control path (1) of a control loop with a state controller (2). In order to achieve optimum adaptation of the state controller (2), it is proposed to adapt control parameters (ri), that is to say the components of the control vector (r), as a function of the respective power level and in a different manner for each parameter (ri). In addition, it is proposed to link the adapted control parameters (ri) by multiplication, not using state variables but only using their change, that is to say using delta state variables ( DELTA xi). <IMAGE>

Description

The invention relates to a control method a process according to the preamble of claim 1.

It is a procedure for a control loop that as a controlled system contains a process that is different Performance levels. At these performance levels the process has a performance level dependent - also in its process parts - different dynamic behavior The dynamic behavior can be determined by a vectorial Differential equation are described. Their solution is then the state vector, the components of which are the state va are variables that are fed to a state controller. In a real process, the state variables are either which is measured or with the help of an observer determined as estimates.

The number of state variables is determined by the order of the determined to be controlled dynamic process. It is iden table with the number of components of the rule vector. Around the optimal control quality for each performance level Take into account the dynamic behavior of the process  to be able to, the individual components of the Regelvek tors, d. H. the control parameters ad be accepted.

In known methods, this adaptation is carried out by multi application of the whole rule vector or of the state reg ler output control signal with the help of an adaptation factor that depends on the level of performance or in general on the loading drive level is dependent.

One such method, from which the invention is based, is in "Control engineering practice" 26th year, 1984, volume 8, Pages 351 to 356. The process considered there is a steam superheater in a steam power plant boiler. Ge regulates the steam temperature. This process also poses a typical, but not the only possible application speed for the method according to the invention such process, the intermediate temperatures as a state variable measurement can only be measured with great effort, one works with an observer who delivers this. Of the The observer is in the aforementioned publication as a series Real-time switching of delay elements siert. To adapt the observer to the dynamic ver Keeping the process going are the time constants of each NEN delay elements depending on the load, i. H. from Steam mass flow, adapted. Furthermore, the observer gain adapted by an external signal. This ex ternal signal is also used around the output signal of the state controller by division. An individual duel adaptation of the components of the rule vector takes place Not.

In the essay by Johannes Mann "Temperature control using state feedback in a fossil fired power plant", IFAC Symposium on Control of power plants and power systems, Munich, March 9 to 11, 1992, Preprints Volume 1, pages 61 to 66 a method for temperature control of a superheater with an observer-state controller combination is also described. In this method, too, it is considered sufficient to adapt the dynamic behavior of the state controller by a single link, shown there in FIG. 1 as a divider. The associated text states that the control parameters fxi remain unchanged when the load changes. Only time constants of the observer are adapted.

The gain factor is the adaptive signal Plant gain is used. Not considered remains the fact that next to the gain factor the time constant of the controlled system also changes depending on the load those who have a say in their dynamic behavior. It can therefore not an optimal or in every operating state Predeterminable dynamic behavior of the control loop achieved become. It will only be a medium one, that is, a compromise ßig setting achieved and thereby a certain robustness achieved.

The invention has for its object a method according to the preamble of claim 1 with respect to a predeterminable optimal dynamics in every performance level improve and thereby the greatest possible robustness aim. The method is intended for systems with or without ob Eighth can be applicable.

This task is characterized by a method with the nenden features of claim 1 in conjunction with the Features of the generic term solved.

The invention proposes the components of the Rule vector individually and depending on performance level adapt. It is also important that the Zu  status variables themselves, but only their changes in each case be multiplied by a component of the rule vector.

The method according to the invention has a number of advantages len. It is readily apparent that the Adap tion of the individual control parameters based on the expansion of the degrees of freedom is a logical proposition Adaptation to the process dynamics is possible. Would you however the individual state variables with load-dependent change multiply th control parameters, so the Transition from one performance level to another large, disruptive changes in the manipulated variables result. The Re The gel process would be restless or even unstable. This The problem is solved according to the invention in that not State variables themselves, but only their changes the control parameters are multiplied.

This measure has the further advantage that the Ver driving is also applicable in control systems that are relatively narrow range of values are limited.

A more detailed description of the invention and of The design is based on execution examples play, which are shown in the drawings.

Show it:

Fig. 1 shows a basic building block for obtaining a signal of amendments from a feed physi rule size, that is multiplied gelvektors with a component of Re,

Fig. 2 shows an embodiment of an arrangement without an observer circuit for performing the control method and

Fig. 3 shows an embodiment of an arrangement with an observer circuit for performing the control method.

In the drawings, reference numerals and signal designations are conditions that correspond in the figures, be the same draws. An index i means in the usual way that meh There are similar signals or components.

Fig. 1 shows a basic block 9 for detecting the change tion of a measured or simulated physical quantity x, which changes by amounts Δx _i ≷0 compared to the initial state, and for multiplication with an individual control parameter r _i . The basic module 9 is used in an adapted form in the exemplary embodiments shown in further figures. It serves to explain the principle used to form a change signal, which can also be used in another context.

The arrangement shown in FIG. 1 contains a first addition point 11 , a second addition point 12 , a first PT1 element 41 with the time constant T, a sixth multiplication point 26 and a first integrator 31 , whose follow-up time constant with the time constant T of the first PT1- Link 41 is identical. A pending physical variable x at the input of the block 9 , or a variable changed by Δx (x + Δx), is di rectly with a positive sign and is also led via the first PT1 element 41 with a negative sign to the first addition point 11 . The output signal from the first addition point 11 is multiplied at the sixth multiplication point 26 with the parameter r _i multi. The output signal of the sixth multiplication point 26 is in each case with a positive sign once di rectly and once via the first integrator 31 to the second addition point 12 , the output signal of which is the adapted change Δx · r _{i to be formed} . Since the output signal of the first addition point 11 was zero in the steady state, the output signal of the sixth multiplication point 26 will not change - if it is in the steady state - if r _{i is} adapted. This means that despite a change in the parameter r _i to r _i + Δr _i, the signal at the output of the second addition point 12 remains unchanged. The new value of r _i + Δr _i only becomes effective in the subsequent transient process.

Fig. 2 shows a first application example of the method according to the Invention, in which directly measured state variables deflect x _i , where i = 1,2. . . , n, a controlled system 1 are available and supplied to a state controller 2 who the. The state controller 2 has integrating behavior, which gives it a second integrator 32 , to which a control deviation (wy) representing the signal E is supplied in an adapted form. This input signal E of the second integrator 32 is formed by subtracting the control variable y from the desired value w at a third addition point 13 and by multiplying its output signal by a control parameter r _{n + 1} at a first multiplication point 21 _{n + 1} . The setpoint w can be set on a setpoint generator 4 . The control parameter r _{n + 1} is formed in a function generator 8 _{n + 1} , the function of which is designated F _{n + 1} . The input signal of the function generator 8 _{n + 1} is an adapting signal Z, which is dependent on the power level.

The arrangement according to FIG. 2 contains several first modules 91 which are modified compared to the basic module 9 shown in FIG. 1 and to which a state variable x 1 to x _{n is} supplied in each case. In the first block 91 , the output signal from the first addition point 11 is multiplied at a first multiplication point 21 1 to 21 _n by the parameter r _i , the parameter r _{i being formed in} each case in a function transmitter 8 _i to which the adapting signal Z is fed.

The output signals A _{i of} the modules 91 are fed to a first summing unit 28 . The output signal of the first totalizer 28 is once directly and once via a third integrator 33, each with a positive sign, to a fourth addition point 14 . The function of the components 14 , 33 corresponds to the function of the components 12 , 31 shown in FIG. 1. The output signal of the fourth addition point 14 represents the sum of the so-called delta state variables adapted by control parameters r _i . Δx _i · r _i . To this signal, the output signal of the second integrator 32 is added at a fifth addition point 15 to form the manipulated variable u .

The combination of delta state variables Δx _i used in the method according to the invention with performance level-dependent control parameters r _i leads to significantly smaller changes in the manipulated variable u when changing the parameter r _i , i = 1 to n, than using the state variable deflect x _i .

To explain this difference, it is assumed that the state variable changes from (x _i ) ₁ to [(x _i ) ₁ + Δx _i ] = (x _i ) ₂ and the component r _i des at the transition of the controlled system from one operating level to the other Change Regelvek tors from (r _i ) ₁ to [(r _i ) ₁ + Δr _i ] = (r _i ) ₂. Then when using the state variable x _i, the change in the i-th input signal of the state controller is:

When using the delta state variables, the change is tion:

It can be seen from the comparison of the two changes that in the method according to the invention the change in the input signals and thereby also the manipulated variable u when the adapted component r _{i changes} by the value (x _i ) ₁ · Δr _i becomes smaller. As a result, the interference effect, which must ultimately be regulated out by the controller as an undesirable fault, is substantially smaller in the control parameter adaptation in the transient state and even zero in the steady state. The adaptation of the control parameters when the power level changes represents a significantly smaller disturbance compared to the natural disturbance variables of the controlled system and can therefore also be implemented. This enables the maximum possible robustness.

Fig. 3 shows a second embodiment wherein the state variables x _i is not as variables need to be formed with the aid of an observer 3 ste available hen, but. Control path 1 is in this example an overheat, in whose steam feed line 10 an injection cooler 50 is inserted, the water supply of which is regulated via a valve 51 with actuator 52 by a state controller 2 with a follower regulator 7 . The slave controller 7 enables a quick response to a change in the steam inlet temperature, which is detected at a measuring point 53 and added at a seventh addition point 17 to the output signal of the state controller 2 .

The state controller 2 contains second building blocks 92 , which are modifications of the building block 9 shown in FIG. 1. The second building blocks 92 are each observed state variables B₁ to B₃ supplied as input variables; ie in the case under consideration n = 3.

Since the state controller 2 is supplied with observed state variables, that is to say not a delta state variable, and the change values of these state variables required for the method according to the invention are determined internally, if the parameters r _{i are} individually adapted, an observer 3 can only be used with a P tracker be the order of the controlled system, z. B. the superheater only increased twice and the exact design of the control vector according to the principle of the separation theorem for the prege bene behavior of the control loop possible. An observer of the delta state variable provides a more complicated structure than that shown in FIG. 3. Such an observer, who would have to include a PI tracker and which, in addition to a superheater model, would also have to include a reference variable model, firstly increases the order of the controlled system more than twice and, secondly, invalidates the separation theory. This means that the design of the control vector for a given dynamic behavior of the control loop (pole default) can only be roughly approximated and the dynamic behavior of the control loop - due to its unnecessarily increased order - becomes sluggish.

The observer 3 used contains second PT1 elements 42 with different time constants T1 to T3. The outlet temperature detected at a second measuring point 54 , ie the controlled variable y leads to a ninth addition point 19 , at which an estimate y formed by the observer 3 is subtracted. The output signal of the ninth addition point 19 is used for tracking the observer via a third multiplication point 23 and an eighth addition point 18 . With h1 to h3 tracking vectors are designated. With V₁ to V₃ reinforcements are called net. The entry temperature detected at the first measuring point is fed to the observer as an input variable.

The time constants of the first PT1 elements 41 in the two-th building blocks 92 are a multiple, in the embodiment, for example, ten times the respective time constants of the second PT1 elements 42 .

The first PT1 elements 41 are each supplied with the input signal G _{i of} the corresponding second PT1 element 42 ; the two PT1 elements 41 , 42 are each connected in parallel on the input side. The output signal of the first PT1-Glie is led to the first addition point 11 with a negative sign. In addition to the one with a positive sign, the simulated state variable B _i , in the case under consideration i = 1, 2, 3, is fed to this first addition point 11, which is tapped at the output of the second PT1 element 42 . The output signal of the first addition point 11 is multiplied at the first multiplication point 21 _i by the adapted control parameter r _i .

The output signal of the first multiplication point 21 is multiplied at a fourth multiplication point 24 by a first factor K11, K12, K13, which is 10/9 in the case under consideration. The result is in each case as an output signal C1, C2, C3 to a second totalizer 29 leads. The output signal of the first multiplication points 21 is also multiplied at fifth multiplication points 25 by second factors K21, K22, K23, the value of which is 1/9 · T _i . The output signals D1, D2, D3 of the fifth multiplication points 25 are guided to a third sum 30 . The signal E is also led to the summing unit 30 , the output of which is connected to a fourth integrator 34 . The outputs of the second summing unit 29 and the fourth integrator 34 are added at a sixth addition point 16 , the output signal of which has a negative sign as the output signal of the state controller 2 to the seventh addition point 17 .

Another exemplary embodiment is a modification not shown in the drawing, in which the building block 91 shown in FIG. 2 is supplied with estimated values formed by an observer instead of the measured state variables x _i .

In addition, the individual control parameters r _i can be determined in the context of an adaptive control, that is to say with the aid of an adaptive algorithm, instead of an adaptation controlled using the signal Z.

Reference list

1 controlled system
2 state controllers
3 observers
4 setpoints
7 slave controllers
8 function formers
9 basic building block
10 steam supply line
11 first addition point
12 second addition point
13 third addition point
14 fourth addition point
15 fifth addition point
16 sixth addition point
17 seventh addition point
18 eighth addition point
19 ninth addition point
21 first multiplication point
22 second multiplication point
23 third multiplication point
24 fourth multiplication point
25 fifth multiplication point
26 sixth multiplication point
28 first totalizer
29 second totalizer
30 third totalizers
31 first integrator
32 second integrator
33 third integrator
34 fourth integrator
41 first PT1 element
42 second PT1 link
50 desuperheaters
51 valve
52 actuator
53 first measuring point
54 second measuring point
91 first module
92 second building block
A _i , C _i , D _i output signals of the modules 91 , 92
B _i observed function variable
E adapted target-actual difference
F _i function of the function generator 8 _i
G _i input signal of the second PT1 element 42
K _i factors
T _i time constants
V _i reinforcements
h _i tracking vectors
r _i control parameters
u manipulated variable
y controlled variable
Estimated value of the controlled variable
w setpoint
Z adapting signal
x _i state variable
Δx _i delta state variable
Δx _i r _i adapted delta state variable

Claims (10)

1. A method for controlling a process which can assume different performance levels, has different dynamic behavior at these performance levels and which forms the controlled system ( 1 ) of a control circuit with a state controller ( 2 ), whereby

• - The state controller ( 2 ) for adapting to the respective dynamic behavior of the controlled system ( 1 ) are supplied with several status variables (x _i ) and
• - an adaptation of the rule vector (r) is also carried out,

characterized in that an optimal adaptation of the state controller ( 2 ) is achieved in that the formation of the manipulated variable (u)

• a) instead of the state variables (x _i ), the changes to these variables (x _i ), that is to say the so-called delta state variables (Δx _i ), are used, and
• b) the delta state variables (Δx _i ) with control parameters (r _i ), ie components of the control vector (r) are multiplied, which are dependent on the performance level and adapted differently depending on the parameter (r _i ), the steady state of the state variables (x _i ) the formed adapted delta state variables (Δx _i · r _i ) remain unchanged when a parameter change (r _i + Δr _i ) occurs.

2. The method according to claim 1, characterized in that all or at least some of the state variables (x _i ) are measured with physical means physical quantities. 3. The method according to claim 1, characterized in that all or at least some of the state variables (x _i ) are formed in an observer ( 3 ). 4. The method according to claim 3, characterized in that an observer ( 3 ) with a P-tracking is used to form state variables (x _i ). 5. The method according to any one of the preceding claims, characterized in that the adapted delta state variables (Δx _i · r _i ) are formed by the following steps:

• a1) a state variable (x _i ) is fed to a first PT1 element ( 41 ), the time constant of which is T, the output signal of which is fed to a first addition point ( 11 ) with a negative sign;
• a2) the state variable (x _i ) also leads directly to the first addition point ( 11 ) with a positive sign;
• b) the output signal of the first addition point ( 11 ) is multiplied at a sixth multiplication point ( 26 ) by an individually adapted control parameter (r _i );
• c) the output signal of the sixth multiplication point ( 26 ) is given a positive sign directly and in addition to that via a first integrator ( 31 ), whose post-setting time constant is T, to a second addition point ( 12 ) whose output signal is the adapted delta to be formed -State variable Δx _i · r _i ).

6. The method according to any one of the preceding claims, characterized in that the individual, ie the performance level and the sub-process - represented by the respective state variable (x _i ) - dependent adaptation of the control parameters (r _i ) with the help of a function generator ( 8 _i ) is performed depending on an adapting signal (Z). 7 The method according to any one of claims 1 to 5, characterized in that the individual adaptation of the control parameters (r _i ) is carried out as part of an adaptive control with the aid of an adaptive algorithm. 8. The method according to claim 6, adapted from the following conversion to an application in which several function variables (x _i to x _n ) are available as physical measured variables:

• a) the function variables (x _i to x _n ) are each fed to a first module ( 91 ), each of which provides an output signal (A _i ) which is an output signal of the first addition point ( 11. ) multiplied by the adapted parameter (r _i ) ) is;
• b) the output signals (A _i ) of the blocks ( 91 ) are summed in a first summing device ( 23 ), and its output signal is fed both directly and via a third integrator ( 33 ) to a fourth addition point ( 14 ), the Output signal is the sum of the delta state variables (Δx _i · r _i ) adapted by control parameters (r _i );
• c) at a fifth addition point ( 15 ) is added to the output signal from the fourth addition point ( 14 ) an adapted setpoint-actual value difference signal to form the manipulated variable (u).

9. The method according to claim 6, adapted by the following from conversion to an application in which a plurality of observed state variables (B _i ) in an observer ( 3 ) with series-connected second PT1 elements ( 42 ), whose time constant T _i , i = 1 to 3, . . . , is to be formed:

• a) the observed state variables (B _i ) are each supplied to a first addition point ( 11 ) in a second module ( 92 );
• b) Input signals (G _i ) of the respective second PT1 circuit ( 42 ) are supplied to first PT1 elements ( 41 ) of the second modules ( 92 ), the time constant being, for. B. 10 · T _i ;
• c) the output signal of the first PT1 element ( 41 ) is led with a negative sign to the first Additi onsstelle ( 11 );
• d) the output signal of the first addition point ( 11 ) is multiplied at a first multiplication point ( 21 _i ) by the adapted parameter (r _i );
• e) the output signal of the first multiplication point ( 21 _i ) is at a fourth multiplication point ( 24 ) with a first factor K11, z. B. 10/9 multiplied to form first output signals (C _i ) of the second building blocks ( 92 );
• f) the output signal of the first multiplication point ( 21 _i ) is also at a fifth multiplication point ( 25 ) with a second factor K12, z. B. 1/9 · T1 multiplied to form second output signals (D _i ) of the second building blocks;
• g) the first output signals (C _i ) are summed with a second totalizer ( 29 ) and led to a sixth addition point ( 16 );
• h) the second output signals (D _i ) are summed with a third totalizer ( 30 ) to which an adapted setpoint-actual value difference signal (E) is also fed;
• i) the output signal of the third totalizer ( 30 ) is integrated with a fourth integrator ( 34 ) and led to the sixth addition point ( 16 ) to form the control signal to be supplied by the state controller ( 2 ).